Data-driven optimization of treatment outcomes in depression. Treatment Selection Idea U. Penn Adam Chekroud Yale University 04/06/16

Transcription

1 Data-driven optimization of treatment outcomes in depression Treatment Selection Idea U. Penn Adam Chekroud Yale University 04/06/16

2 $9.6 Billion U.S. spending on antidepressants 70% do not fully recover from initial treatment

3

4 OUTLINE 1. Objective: Prospectively identify antidepressant responders 2. Statistical modeling procedures 3. Performance generalizes to independent clinical trial 4. Implementation at scale 5. Recent progress

5 Articles Cross-trial prediction of treatment outcome in depression: a machine learning approach Adam Mourad Chekroud, Ryan Joseph Zotti, Zarrar Shehzad, Ralitza Gueorguieva, Marcia K Johnson, Madhukar H Trivedi, Tyrone D Cannon, John Harrison Krystal, Philip Robert Corlett Summary Background Antidepressant treatment efficacy is low, but might be improved by matching patients to interventions. At present, clinicians have no empirically validated mechanisms to assess whether a patient with depression will respond to a specific antidepressant. We aimed to develop an algorithm to assess whether patients will achieve symptomatic remission from a 12-week course of citalopram. Methods We used patient-reported data from patients with depression (n=4041, with 1949 completers) from level 1 of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D; ClinicalTrials.gov, number NCT ) to identify variables that were most predictive of treatment outcome, and used these variables to train a machine-learning model to predict clinical remission. We externally validated the model in the escitalopram treatment group (n=151) of an independent clinical trial (Combining Medications to Enhance Depression Outcomes [COMED]; ClinicalTrials.gov, number NCT ). Findings We identified 25 variables that were most predictive of treatment outcome from 164 patient-reportable variables, and used these to train the model. The model was internally cross-validated, and predicted outcomes in the STAR*D cohort with accuracy significantly above chance (64 6% [SD 3 2]; p<0 0001). The model was externally validated in the escitalopram treatment group (N=151) of COMED (accuracy 59 6%, p=0.043). The model also performed significantly above chance in a combined escitalopram-buproprion treatment group in COMED (n=134; accuracy 59 7%, p=0 023), but not in a combined venlafaxine-mirtazapine group (n=140; accuracy 51 4%, p=0 53), suggesting specificity of the model to underlying mechanisms. Lancet Psychiatry 2016 Published Online January 20, S (15)00471-X See Online/Comment S (15) Department of Psychology (A M Chekroud MSc, Z Shehzad MSc, M K Johnson PhD, T D Cannon PhD), Department of Biostatistics (R Gueorguieva PhD), and Department of Psychiatry (T D Cannon, J H Krystal MD, P R Corlett PhD), Yale University, New Haven, CT, USA; Capital One, McLean, VA, USA (R J Zotti BSc); Department of Psychiatry, UT Southwestern,

6 STAR*D - DESIGN Largest clinical trial of MDD (N= 4,041) Archival dataset - freely-available from NIMH Prospective, randomized clinical trial; realistic sampling Level 1: 12-week SSRI treatment

7 STAR*D - DESIGN Amongst Level 1 treatment completers, label patients that reached remission (QIDS 5) STAR*D Cohort Citalopram responder Citalopram non-responder

8 STAR*D - DATA Wide Feature Set Extract all readily-available data Demographic Diagnostic criteria (DSM-IV) Questionnaire responses about symptoms Psychiatric History Family history Chekroud et al. (2016) Lancet Psychiatry

9 FEATURE SELECTION Wide Feature Set Elastic Net Regularization Best 25 Questions Zou & Hastie, J.R. Stat. Soc. (2005)

10 FEATURE SELECTION Wide Wide Feature Feature SetSet Elastic Net Regularization Best 25 Questions 1!"#!!! (!!,!) ℝ 2!! (!!!!!!!!)!! +!)!! +![(1!)!!! /2! +!!! /2 +!!! ],!!!! Ordinary Least Squares Ridge/L2 Lasso/L1 Zou & Hastie, J.R. Stat. Soc. (2005)

11 BEST PREDICTIVE ITEMS β Item Initial Depressive Severity (QIDS-C) Currently Employed Feeling Restless (Psychomotor Agitation) Low Energy Levels (Fatigue) Ethnicity = Black/African American Chekroud et al. (2016) Lancet Psychiatry

12 Amongst Level 1 treatment completers, label patients that reached remission (QIDS 5) TRAIN PREDICTIVE MODEL Citalopram responder Citalopram non-responder Train GBM using only 25 features Relative Variable Importance All Information Available Before Treatment Elastic Net Regularization Best 25 Predictors Chekroud et al. (2016) Lancet Psychiatry

13 MODEL PERFORMANCE Test-fold Performance Gradient Boosting Machine Accuracy 65% p-value [acc>nir] 9.8 x10-33 Sensitivity 63% PPV 64% PPV = Positive Predictive Value Chekroud et al. (2016) Lancet Psychiatry

14 WHAT NEXT? 1. Objective: Prospectively identify antidepressant responders 2. Statistical modeling procedures 3. Performance generalizes to independent clinical trial 4. Implementation at scale 5. Recent progress

15 STAR*D Sequenced Treatment Alternatives to Relieve Depression COMED Combination Medication to Enhance Depression Outcomes A) External Generalization of STAR*D Model CIT only N = 1,949 Acc 60% * Sens 49%, Spec 71% ESC + PLACEBO N = 151 Acc 65% *** AUC 0.70 Sens 63% Spec 66% Acc 60% * Sens 56%, Spec 63% Acc 51% [n.s] Sens 39%, Spec 65% ESC + BUP N = 134 VEN + MIR N = 140 Chekroud et al. (2016) Lancet Psychiatry

16 WHAT NEXT? 1. Objective: Prospectively identify antidepressant responders 2. Statistical modeling procedures 3. Performance generalizes to independent clinical trial 4. Implementation at scale 5. Recent progress

17

18 Progress WE RECOMMEND

19

20 OUTLINE 1. Objective: Prospectively identify antidepressant responders 2. Statistical modeling procedures 3. Performance generalizes to independent clinical trial 4. Implementation at scale 5. Recent progress

21 Antidepressant Drug Effects and Depression Severity A Patient-Level Meta-analysis Jay C. Fournier, MA Robert J. DeRubeis, PhD Steven D. Hollon, PhD Sona Dimidjian, PhD Jay D. Amsterdam, MD Richard C. Shelton, MD Jan Fawcett, MD ANTIDEPRESSANT MEDICATION (ADM) represents the current standard of treatment for major depressive disorder (MDD). 1 Antidepressant medication has Context Antidepressant medications represent the best established treatment for major depressive disorder, but there is little evidence that they have a specific pharmacological effect relative to pill placebo for patients with less severe depression. Objective To estimate the relative benefit of medication vs placebo across a wide range of initial symptom severity in patients diagnosed with depression. Data Sources PubMed, PsycINFO, and the Cochrane Library databases were searched from January 1980 through March 2009, along with references from meta-analyses and reviews. Study Selection Randomized placebo-controlled trials of antidepressants approved by the Food and Drug Administration in the treatment of major or minor depressive disorder were selected. Studies were included if their authors provided the requisite original data, they comprised adult outpatients, they included a medication vs placebo comparison for at least 6 weeks, they did not exclude patients on the basis of a placebo washout period, and they used the Hamilton Depression Rating Scale Differential efficacy of escitalopram and nortriptyline on dimensional measures of depression Rudolf Uher, Wolfgang Maier, Joanna Hauser, Andrej Marušič, Christine Schmael, Ole Mors, Neven Henigsberg, Daniel Souery, Anna Placentino, Marcella Rietschel, Astrid Zobel, Monika Dmitrzak-Weglarz, Ana Petrovic, Lisbeth Jorgensen, Petra Kalember, Caterina Giovannini, Mara Barreto, Amanda Elkin, Sabine Landau, Anne Farmer, Katherine J. Aitchison and Peter McGuffin

22 DEPRESSIVE SEVERITY 1. Low mood 2. Insomnia 3. Feelings of guilt/worthlessness 4. Cognitive impairments 5. Weight/appetite changes 6. Suicidal ideation Chekroud, Gueorguieva, Krumholz, Trivedi, Krystal & McCarthy. (in prep)

23 SUB-SYMPTOM STRUCTURE QIDS Mid nocturnal insomnia Sleep onset insomnia Early morning insomnia Energy/Fatigability Concentration/decision making Involvement Mood (sad) Outlook (self) Psychomotor Agitation Psychomotor Slowing Suicidal ideation Hypersomnia Chekroud, Gueorguieva, Krumholz, Trivedi, Krystal & McCarthy. (in prep)

24 REPLICABLE STRUCTURE? Waking during the night Difficulties falling asleep Waking up too early Low energy levels Feeling sad SLEEP Poor concentration/decision making COGNITIVE / AFFECTIVE No interest in people/activities Low self-worth Restlessness/ Fidgeting Slowed movement and thinking ATYPICAL Suicidal ideation Cannot get out of bed Waking during the night Difficulties falling asleep Waking up too early Low energy levels Poor concentration/decision making No interest in people/activities Feeling sad SLEEP COGNITIVE / AFFECTIVE Low self-worth Restlessness/ Fidgeting Slowed movement and thinking ATYPICAL Suicidal ideation Cannot get out of bed STAR*D COMED Chekroud, Gueorguieva, Krumholz, Trivedi, Krystal & McCarthy. (in prep)

25 SUB-SYMPTOM TRAJECTORIES 2.0 Citalopram Average Severity Sleep/insomnia Core Depressive Atypical Duration of Treatment (Weeks) Chekroud, Gueorguieva, Krumholz, Trivedi, Krystal & McCarthy. (in prep)

26 SUB-SYMPTOM TRAJECTORIES 2.0 Citalopram 2.0 Escitalopram + Placebo Sleep/insomnia Average Severity Average Severity Core Depressive Atypical Duration of Treatment (Weeks) Duration of Treatment (Weeks) Escitalopram + Buproprion Venlafaxine + Mirtazapine Average Severity Average Severity Duration of Treatment (Weeks) Duration of Treatment (Weeks) Chekroud, Gueorguieva, Krystal & McCarthy. (in prep)

27 SUB-SYMPTOM TRAJECTORIES 12 week symptom cluster response trajectories QIDS Atypical Core Sleep 1.5 Average Severity Duration of Treatment (Weeks) Citalopram Escitalopram...Bupropion Escitalopram...Placebo Venlafaxine...Mirtazapine Chekroud, Gueorguieva, Krystal & McCarthy. (in prep)

28 SUB-SYMPTOM TRAJECTORIES Chekroud, Gueorguieva, Krystal & McCarthy. (in prep)

29 SUB-SYMPTOM TRAJECTORIES week symptom cluster response trajectories HAM D Atypical Core Depressive Sleep 1.2 Average Severity Duration of Treatment (Weeks) Duloxetinehigh Duloxetinelow Escitalopram Fluoxetine Paroxetine Placebo Chekroud, Gueorguieva, Krystal & McCarthy. (in prep)

30 SUMMARY 1. Statistical models might guide medication choice in Depression 2. Statistical models are not the end goal! 3. Extensions: A. Finer predictions? Better predictions? B. Other treatments, other mental illnesses

31 SUMMARY Antidepressants don t work on all symptoms Core symptoms yes; sleep symptoms sometimes, atypical not much Pick the most effective AD for *your symptoms* core/atypical ~ mg BID duloxetine, 20mg paroxetine QD sleep ~ mg venlafaxine mirtazapine QD Streamline and optimize care pathways for mental illness efficient screening ~ ipad in waiting room treatment selection ~ predictive modeling outcome tracking ~ is it working!?! (& more data, more models, better tmt)

32 ACKNOWLEDGEMENTS Gregory McCarthy John Krystal Ralitza Gueorguieva Marcia Johnson Phil Corlett Harlan Krumholz Tyrone Cannon Madhukar Trivedi